GISTs are the most common mesenchymal tumors associated with the gastrointestinal tract with a global incidence of 10–15 cases per million (Joensuu et al., 2002; Mucciarini et al., 2007; Serrano and George, 2014; Zhang et al., 2021). Most GISTs are genetically driven by KIT and PDGFRA mutations. Before the 21st century, GISTs were also known as gastrointestinal stromal sarcomas and demonstrated a high rate of primary resistance to chemotherapy and radiation that led to a poor overall prognosis (Hirota et al., 1998; Kim and Zalupski, 2011). However, imatinib (the first TKI approved by the US Food and Drug Administration (FDA) for therapy of GISTs) has revolutionized the therapeutic approach to GISTs and improved the prognosis of patients with advanced GISTs from a pre-imatinib 2-year overall survival (OS) rate of 25% to approximately 70–80% (Demetri et al., 2002; Verweij et al., 2004; Blanke et al., 2008; Le Cesne et al., 2010). Although imatinib resulted in an unprecedented efficacy and tolerance in the GISTs with a median progression-free survival (mPFS) of almost 2 years, most patients acquired secondary mutations and approximately 15% of the patients exhibited primary resistance to imatinib (Antonescu et al., 2005; Joensuu, 2006; Antonescu, 2008; Demetri, 2011). Secondary mutations that mainly occur in the ATP-bounding domain or activation loop blocked the binding of some TKIs and resulted in tumor progression (Liegl et al., 2008; Namlos et al., 2018).

Sunitinib, a multi-targeted TKI that inhibits KIT, PDGFRA, and vascular endothelial growth factor receptors (VEGFRs), is approved for the treatment of advanced GIST after the failure of imatinib (Demetri et al., 2006). In a phase III trial of advanced GISTs after the failure of imatinib, sunitinib displayed an mPFS of 22.9 weeks (Demetri et al., 2006; Demetri et al., 2012).

Even after receiving this second-line treatment, most patients continue to experience disease progression; these patients are considered refractory cases and require third-line or fourth-line treatment (regorafenib and ripretinib). The third-line (regorafenib) treatment demonstrated an mPFS of 4.8 months; this discouraging outcome probably indicates incomplete suppression of secondary resistant mutations.

Presently, there is no consensus on refractory therapies despite the National Comprehensive Cancer Network (NCCN) guidelines (version 1. 2021) that recommend sorafenib, dasatinib, nilotinib, pazopanib, and everolimus plus TKI as potential options for refractory cases.

Consequently, it is essential to simultaneously compare the efficacy and tolerance of multiple regimens for refractory cases. Recently, several novel TKIs, such as ripretinib, pazopanib, and pimitespib, have displayed promising superiority over the placebo (Mir et al., 2016; Blay et al., 2020; Kang et al., 2021). In this network review, we estimate and rank the efficacy and tolerance of all possible therapies for patients with advanced refractory GIST, which were previously treated with at least first- and second-line regimens, using a Bayesian network model.

To our knowledge, this is the first Bayesian network meta-analysis to assess the comparative efficacy and tolerance of third-line or over third-line therapies against GIST. Zhang et al. performed a Bayesian network meta-analysis on targeted therapies after the prior failure of TKIs (Zhang et al., 2021). However, in this study, we aimed at third-line and over third-line therapies and analyzed more beneficial clinical outcomes using random effects models.

In the network meta-analysis of seven RCTs consisting of a total of 1300 patients with refractory GIST, we simultaneously analyzed eight treatment nodes including seven TKIs and one placebo/BSC reference. We found that ripretinib was superior to other regimens with respect to both PFS and DCR, which displayed higher SUCRA values, especially in the fourth-line therapy subgroup. It was also determined that pimitespib and nilotinib had better tolerance.

This indicated that imatinib could achieve an overall response rate (ORR) of >50% (CR: 5%; PR: 48.3%) (Demetri et al., 2002; van Oosterom et al., 2002; Verweij et al., 2004; Kubota, 2006). Nevertheless, none of the eligible trials for this systematic review reached CRs, and only a few PRs were reported. The ORR in several studies might be 0, which had comparatively no significance. For this reason, we selected DCR instead of ORR as an outcome to evaluate the efficacy of the treatments.

Ripretinib, a broad-spectrum KIT and PDGFRA switch-control TKI, was the most effective agent in improving PFS, OS, and DCR (Dhillon, 2020). The heterogeneity of drug-resistant KIT mutations has been a major challenge in the treatment of GISTs; the presence of a broad inhibitor with outstanding effectiveness is inspiring (Smith et al., 2019).

Ripretinib is approved for the treatment of adult patients with advanced GIST who have received prior treatment with ≥3 TKIs including imatinib (Lostes-Bardaji et al., 2021). Recently, a phase III study of ripretinib versus sunitinib for the treatment of advanced GISTs after the failure of imatinib (INTRIGUE) was completed (Bauer et al., 2022). However, ripretinib did not meet the primary end point of superiority in the PFS over sunitinib (8 months vs. 8.3 months), whereas ORR was higher for patients receiving ripretinib in the KIT exon 11 population and ripretinib has a more favorable safety profile (Bauer et al., 2022). Although ripretinib did not seem to defeat sunitinib regarding second-line therapy for GISTs, it still performed better in this network meta-analysis, which suggests ripretinib as a high priority in fourth-line therapy.

More specifically, regorafenib was best in improving PFS in the subgroup analysis of only third-line therapy. Regorafenib is a multi-target inhibitor for KIT, platelet-derived growth factor receptors (PDGFRs), VEGFR, fibroblast growth factor receptors, RET, and BRAF and is suggested for the third-line therapy for GISTs based on the guidelines (Nishida et al., 2016). The direct comparative trial VOYAGER (Kang et al., 2021) (avapritinib vs. regorafenib) also recently failed. Although traditional TKIs, such as sunitinib and regorafenib, did not show an ideal effectiveness as novel TKIs in the trial stage, they are still reliable for clinical administration.

The available research data show that the sensitivity of different genetic subtypes to drugs is inconsistent in GIST. Clinical activity of sunitinib after imatinib failure is significantly influenced by both primary and secondary mutations in the predominant pathogenic kinase, which has implications for optimization of the treatment of patients with GISTs, and regorafenib has also shown variable sensitivity to multiple resistant tumor cell clones driven by additional mutations in KIT or PDGFRA (Heinrich et al., 2008; Yeh et al., 2017; Blay et al., 2021). Avapritinib has excellent efficacy in the treatment of unresectable or metastatic PDGFRA D842V mutant GISTs (Heinrich et al., 2020). Ripretinib, as a broad-spectrum KIT and PDGFRA switch-control TKI, showed excellent efficacy against various primary and secondary mutation types in the fourth-line treatment of GIST, although it can be observed from the genotyping data of the INTRIGUE study that efficacy against different genetic subtypes is still different (Blay et al., 2020; Bauer et al., 2022). Precise targeted therapy is a very important direction for the future treatment of GIST, and more high-quality studies of new agents are expected. We analyzed the PFS outcome of the 11-exon mutation and 9 exon mutation subgroups. The results of a few studies consisted of survival status correlated with genomic analysis, and only 2 studies with 3 treatment nodes were included in the network meta-analysis. This limited sample size might reduce the persuasiveness of the comparison. However, along with the unsatisfying results of the INTRIGUE trial, which temporarily denied an all different mutation types, the next breakthrough of a new drug should focus on specific gene types. In this case, we believe that more gene-specific prognostic survival data would be particularly emphasized in the coming clinical trials. In our vision, gene-mutation-specific survival is instructive.

In this network meta-analysis, we noticed a new drug (pimitespib, TAS-116) administered for the treatment of GISTs (Kurokawa et al., 2022). Pimitespib is an oral heat shock protein 90 (HSP90) inhibitor that is used for the treatment of T-cell leukemia or colorectal cancer (Kawazoe et al., 2021; Ikebe et al., 2022). Pimitespib also has the potential of improving OS and shows a high tolerance in our review.

This study has some limitations.

First, this study was limited to estimations that were based on the data availability. As the assessment of OS required a long period of time, Mir et al. (2016) and Kang et al. (2021) did not report the HR of OS because it was not reached. In terms of AEs, the total number of grade 3–5 AEs were not present in studies by Mir et al. (2016) and Blay et al. (2020), and we could not calculate it by the accumulation either because patients might suffer from more than one pattern of AE; therefore, pazopanib and ripretinib were not included in the comparison even though ripretinib was reported to have a more favorable safety profile compared with sunitinib (Bauer et al., 2022).

Second, because of the low number of included studies and each comparison between treatments consisted of only one study, the statistical power might be relatively lower in the case of heterogeneity. However, no obvious heterogeneity was found in the present meta-analysis.

Third, this network meta-analysis was performed based on the summary or aggregated data instead of individual data, which is common in meta-analyses. However, we did not have access to enough detailed data for the subgroup analysis.

Fourth, publication bias actually did exist in the present network meta-analysis. It is natural that positive outcomes indicating novel and effective drugs are more likely to be published. A few clinical trials on new drugs which were non-effective in a prior stage could be withdrawn. Additionally, RCTs focused on GISTs were limited, which could be another source of publication bias contributing to the rarity.

Fifth, no loop structure was constructed in the network relationship plot. For this reason, we did not perform a loop inconsistency test to assess the inconsistencies in this network meta-analysis. We hope that more direct comparisons of the agents can be carried out.

Sixth, though combined as treatments for refractory GIST, the third-line and fourth-line therapies could still be under different clinical contexts. We performed a subgroup analysis to represent detailed results after the comparison.

We verified that regorafenib and ripretinib could be more reliable regimens for third-line and over-third-line therapies, respectively. In pooled analysis, ripretinib expressed priority in improving PFS, OS, and DCR. Both nilotinib and pimitespib presented favorable tolerance, whereas their efficacy still needs to be verified. Pimitespib is a novel potential agent for the treatment of GISTs, and more details are expected. Using an inhibitor for a specific gene mutation type may be a new trend in the future. We hope that more high-quality studies of new agents can bring new information to light in the area of precise targeted therapy for GISTs.